ARTICLE https://doi.org/10.1038/s41467-021-24798-y OPEN PRMT1-dependent regulation of RNA metabolism and DNA damage response sustains pancreatic ductal adenocarcinoma ✉ Virginia Giuliani 1 , Meredith A. Miller1,17, Chiu-Yi Liu1,17, Stella R. Hartono 2,17, Caleb A. Class 3,13, Christopher A. Bristow1, Erika Suzuki1, Lionel A. Sanz2, Guang Gao1, Jason P. Gay1, Ningping Feng1, Johnathon L. Rose4, Hideo Tomihara4,14, Joseph R. Daniele1, Michael D. Peoples1, Jennifer P. Bardenhagen5, Mary K. Geck Do5, Qing E. Chang6, Bhavatarini Vangamudi1,15, Christopher Vellano1, Haoqiang Ying 7, Angela K. Deem1, Kim-Anh Do3, Giannicola Genovese4,8, Joseph R. Marszalek1, Jeffrey J. Kovacs1, Michael Kim9, 1234567890():,; Jason B. Fleming9,16, Ernesto Guccione10, Andrea Viale4, Anirban Maitra 11, M. Emilia Di Francesco5, Timothy A. Yap 12, Philip Jones 5, Giulio Draetta 1,4,5, Alessandro Carugo 1, Frederic Chedin 2 & ✉ Timothy P. Heffernan 1 Pancreatic ductal adenocarcinoma (PDAC) is an aggressive cancer that has remained clini- cally challenging to manage. Here we employ an RNAi-based in vivo functional genomics platform to determine epigenetic vulnerabilities across a panel of patient-derived PDAC models. Through this, we identify protein arginine methyltransferase 1 (PRMT1) as a critical dependency required for PDAC maintenance. Genetic and pharmacological studies validate the role of PRMT1 in maintaining PDAC growth. Mechanistically, using proteomic and transcriptomic analyses, we demonstrate that global inhibition of asymmetric arginine methylation impairs RNA metabolism, which includes RNA splicing, alternative poly- adenylation, and transcription termination. This triggers a robust downregulation of multiple pathways involved in the DNA damage response, thereby promoting genomic instability and inhibiting tumor growth. Taken together, our data support PRMT1 as a compelling target in PDAC and informs a mechanism-based translational strategy for future therapeutic development. Statement of significance PDAC is a highly lethal cancer with limited therapeutic options. This study identified and characterized PRMT1-dependent regulation of RNA metabolism and coordination of key cellular processes required for PDAC tumor growth, defining a mechanism-based transla- tional hypothesis for PRMT1 inhibitors. *A list of author affiliations appears at the end of the paper. NATURE COMMUNICATIONS | (2021) 12:4626 | https://doi.org/10.1038/s41467-021-24798-y | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24798-y ancreatic cancer is projected to become the second leading in vivo functional genomics platform, to identify epigenetic vul- Pcause of tumor-related deaths by 20301 due to its increasing nerabilities across a panel of fully annotated patient-derived incidence, poor overall 5-year survival rate, and limited PDAC models with known engraftment efficiency (Supplemen- therapeutic options. Pancreatic ductal adenocarcinoma (PDAC) tary Fig. 1a). These models were confirmed to harbor KRAS and accounts for more than 80% of pancreatic cancer cases2. Next- TP53 mutations in addition to lower frequency mutations in generation sequencing (NGS) studies have uncovered the mole- known PDAC-associated genes (Supplementary Data 1). Using a cular mechanisms that contribute to PDAC pathogenesis, previously described approach5, we interrogated these models including oncogenic mutations in KRAS and inactivating muta- with a high-complexity lentiviral library targeting 237 epigenetic tions in the tumor suppressor genes TP53, SMAD4, and regulators (10 shRNA/gene) (Fig. 1a). Analysis of shRNA den- CDKN2A2. Unbiased sequencing studies have also identified sities and fold change in tumors confirmed depletion of shRNAs lower-frequency mutational events impacting core biological that targeted genes essential for in vivo tumor growth compared networks, which include DNA damage repair (DDR), axon gui- to the cell population pre-transplantation (Supplementary dance, and epigenetic regulation3,4. Fig. 1b). Adequate separation of positive (RPL30, PSMA1) and Given the penetrance of loss of-function epigenetic mutations, negative (LUC) controls was confirmed (Supplementary Fig. 1c). such as alterations in histone-modifying enzymes and chromatin By applying stringent thresholds (redundant shRNA activity remodeling complexes3,4, we aimed to identify specific epigenetic (RSA) LogP ≤ −1.5 in at least one PDX and FDR ≤ 0.3), we vulnerabilities that can be exploited therapeutically. We thus captured individual (Supplementary Data 2) and common epi- employed our in vivo target discovery platform, PILOT (Patient- genetic vulnerabilities across PDX models (Fig. 1b and Supple- based In vivo Lethality to Optimize Treatment),5 to systematically mentary Data 3). This analysis confirmed known synthetic lethal uncover epigenetic dependencies in a panel of patient-derived interactions, including ARID1B depletion in ARID1A-mutated xenograft (PDX) models of PDAC. Through this unbiased PDAC tumors, which have been previously demonstrated in approach, we identified protein arginine methyltransferase 1 other tumor types15 (Supplementary Fig. 1d). These findings thus (PRMT1) as a top-scoring hit and a novel genetic vulnerability validate our ability to properly associate gene essentiality with in PDAC. relevant molecular features. Arginine methylation is a common post-translational modifica- Through this effort, five genes emerged as common lethality tion that regulates multiple cellular processes6‒8. Protein arginine factors in all PDAC PDX models (Fig. 1c). Two of these, PHF5A methyltransferases (PRMTs), the only enzymes that mediate this and SMC2, were consistent with our previous work in PDAC5. reaction, catalyze the transfer of a methyl group from S-adenosyl KIF11 encodes a motor protein required for proper spindle methionine (SAM) to the arginine residues of histone and non- assembly and has been reported as a top-scoring hit in genetic histone proteins9. All nine PRMT family members mediate the screens of cell fitness16. CHD4 encodes a chromodomain- addition of one methyl group to one of the guanidine nitrogens of containing protein that catalyzes ATP-dependent nucleosome arginine, generating mono-methylarginine (MMA). PRMT family remodeling and was previously identified as a vulnerability in members are classified into three types based on the final methy- breast cancer17. We were particularly interested in PRMT1, larginine product that is generated. Specifically, Type I enzymes which encodes for an arginine N-methyltransferase that was a catalyze the addition of a second methyl group to the same nitro- top-scoring gene across all PDAC models (Supplementary gen, producing asymmetric di-methylarginine (ADMA); Type II Fig. 1e). PRMT1 belongs to a druggable family of enzymes, enzymes methylate additional guanidine nitrogen, producing sym- and PRMT type I inhibitors are currently under clinical metric dimethylarginine (SDMA); and the Type III enzyme, investigation12,13. PRMT1 has been documented to drive pro- PRMT7, solely catalyzes MMA6,9. Interestingly, dysregulation of tumorigenic events in multiple tumor types7,10,including arginine methylation has been increasingly associated with PDAC18‒20. However, there remains great interest in defining cancer10,11,andPRMTshavethusgarneredsignificant interest as the mechanisms by which PRMT1 contributes to PDAC therapeutic targets. Accordingly, several agents targeting PRMTs pathogenesis. have been developed, with PRMT Type I and PRMT5 selective inhibitors currently under clinical investigation12,13. In this study, we identify PRMT1 as a novel vulnerability in PRMT1 is a critical dependency in PDAC. PRMT1 is the pri- PDAC PDX models and demonstrate context-specific depen- mary methyltransferase that catalyzes asymmetric dimethylation dency using both genetic and pharmacological approaches. of arginine residues (ADMA). Accordingly, global depletion of PRMT1 is the predominant Type I enzyme responsible for more ADMA with a concurrent increase of global arginine mono- than 85% of ADMA14, which regulates a variety of cellular methylation (MMA) is indicative of efficient PRMT1 inhibition14. processes6,7. However, the mechanistic basis of PRMT1 depen- To validate the dependency of PDAC on PRMT1 expression, we dency within specific tissue and genetic contexts, including engineered PATC53 cells, a patient-derived model used in the PDAC, remains poorly understood. Thus, we leveraged ortho- PILOT screening, with two independent, doxycycline-inducible gonal proteomic and transcriptomic approaches to elucidate the shRNAs targeting PRMT1 or a luciferase non-targeting (NT) molecular mechanism underlying the role of ADMA in PDAC. control shRNA (Fig. 1d). Using methylarginine-specific anti- Our studies demonstrated that ADMA is required for faithful bodies (Supplementary Fig. 2a), we observed decreased levels of RNA processing as well as for the expression of multiple genes ADMA and accumulation of MMA in PRMT1-depleted cells and pathways required for supporting PDAC maintenance and compared to NT shRNA and no-doxycycline negative controls. genome stability, including cell cycle and DDR networks. We also The banding pattern on Western blots highlighted the large confirmed the requirement of PRMT1 to maintain in vivo growth number of protein targets post-transcriptionally modified by of PDAC PDXs, suggesting a mechanism-based translational PRMT1 (Fig. 1d, MMA; Supplementary Fig. 2b, ADMA). PRMT1 hypothesis for the clinical development of